Video recognition is an increasingly popular research area with applications in security, human behavior analysis, education, entertainment and different business tasks. Because of the complexity of most of the video data, this area needs new scientific methods before we can benefit from it in practice. In this presentation, we will consider the most advanced methods including hand-crafted features based on optical flow and learned features like convolutional neural networks. Some of the previous results were reproduced and improved in my works. Finally, a new method based on our recent developments in the image recognition area along with its results will be present and discussed.
Ort: INB Seminarraum